How to use

Logarithmic

in

cern.jet.random.tdouble

/**
 * Returns a random number from the distribution.
 */
public static double staticNextDouble(double p) {
  synchronized (shared) {
    return shared.nextDouble(p);
  }
}

/**
 * Constructs a Logarithmic distribution.
 */
public Logarithmic(double p, DoubleRandomEngine randomGenerator) {
  setRandomGenerator(randomGenerator);
  setState(p);
}

  /**
   * Sets the uniform random number generated shared by all <b>static</b>
   * methods.
   * 
   * @param randomGenerator
   *            the new uniform random number generator to be shared.
   */
  private static void xstaticSetRandomGenerator(DoubleRandomEngine randomGenerator) {
    synchronized (shared) {
      shared.setRandomGenerator(randomGenerator);
    }
  }
}

  /**
   * Sets the uniform random number generated shared by all <b>static</b>
   * methods.
   * 
   * @param randomGenerator
   *            the new uniform random number generator to be shared.
   */
  private static void xstaticSetRandomGenerator(DoubleRandomEngine randomGenerator) {
    synchronized (shared) {
      shared.setRandomGenerator(randomGenerator);
    }
  }
}

/**
 * Constructs a Logarithmic distribution.
 */
public Logarithmic(double p, DoubleRandomEngine randomGenerator) {
  setRandomGenerator(randomGenerator);
  setState(p);
}

/**
 * Returns a random number from the distribution.
 */
public static double staticNextDouble(double p) {
  synchronized (shared) {
    return shared.nextDouble(p);
  }
}

/**
 * Returns a random number from the distribution; bypasses the internal
 * state.
 */
public double nextDouble(double a) {
  /***********************************************************************
   * * Logarithmic Distribution - Inversion/Transformation * *
   * ***************************************************************** *
   * The algorithm combines Inversion and Transformation. * It is based on
   * the following fact: A random variable X from * the Logarithmic
   * distribution has the property that X for fixed * Y=y is Geometric
   * distributed with P(X=x|Y=y)=(1-y)*y^(x-1) (*) * where Y has
   * distribution function F(y)=ln(1-y)/ln(1-p). * So first random numbers
   * y are generated by simple Inversion, * then k=(long int)
   * (1+ln(u)/ln(y)) is a Geometric random number * and because of (*) a
   * Logarithmic one. * To speed up the algorithm squeezes are used as
   * well as the * fact, that many of the random numbers are 1 or 2
   * (depending on * special circumstances). * On an IBM/PC 486 optimal
   * performance is achieved, if for p<.97 * simple inversion is used and
   * otherwise the transformation. * On an IBM/PC 286 inversion should be
   * restricted to p<.90. * *
   * ***************************************************************** *
   * FUNCTION: - lsk samples a random number from the * Logarithmic
   * distribution with * parameter 0 < p < 1 . * REFERENCE: - A.W. Kemp
   * (1981): Efficient generation of * logarithmically distributed
   * pseudo-random * variables, Appl. Statist. 30, 249-253. * SUBPROGRAMS:
   * - drand(seed) ... (0,1)-Uniform generator with * unsigned long
   * integer *seed. * *
   **********************************************************************/
  double u, v, p, q;

/**
 * Returns a random number from the distribution; bypasses the internal
 * state.
 */
public double nextDouble(double a) {
  /***********************************************************************
   * * Logarithmic Distribution - Inversion/Transformation * *
   * ***************************************************************** *
   * The algorithm combines Inversion and Transformation. * It is based on
   * the following fact: A random variable X from * the Logarithmic
   * distribution has the property that X for fixed * Y=y is Geometric
   * distributed with P(X=x|Y=y)=(1-y)*y^(x-1) (*) * where Y has
   * distribution function F(y)=ln(1-y)/ln(1-p). * So first random numbers
   * y are generated by simple Inversion, * then k=(long int)
   * (1+ln(u)/ln(y)) is a Geometric random number * and because of (*) a
   * Logarithmic one. * To speed up the algorithm squeezes are used as
   * well as the * fact, that many of the random numbers are 1 or 2
   * (depending on * special circumstances). * On an IBM/PC 486 optimal
   * performance is achieved, if for p<.97 * simple inversion is used and
   * otherwise the transformation. * On an IBM/PC 286 inversion should be
   * restricted to p<.90. * *
   * ***************************************************************** *
   * FUNCTION: - lsk samples a random number from the * Logarithmic
   * distribution with * parameter 0 < p < 1 . * REFERENCE: - A.W. Kemp
   * (1981): Efficient generation of * logarithmically distributed
   * pseudo-random * variables, Appl. Statist. 30, 249-253. * SUBPROGRAMS:
   * - drand(seed) ... (0,1)-Uniform generator with * unsigned long
   * integer *seed. * *
   **********************************************************************/
  double u, v, p, q;